This specification relates to parallel processing large datasets of non-scalar data.
Machine learning models receive input and generate output based on the received input and on values of model parameters. These models often need to be trained using received inputs which are very large datasets.